How it works
From an address to an inspectable investigation.
Memoir investigates a public Ethereum or Base address in one workspace. It asks independent public sources for account facts, assets, intelligence signals, relationships, and chronological activity; keeps missing coverage visible; and lets the user inspect the evidence before optionally asking an AI model to explain a selected snapshot.
The investigation flow
- Enter one public address. Memoir accepts an Ethereum or Base
0xaddress. It never asks for a seed phrase, private key, signing permission, or wallet connection. - Read sources independently. Account history, holdings, protocol state, public labels, sanctions data, and other signals can succeed or fail separately.
- Inspect one workspace. Account, assets and positions, public signals, direct relationships, and chronological activity remain separate modules in the same address record.
- Select what matters. Users can add timeline records and structured workspace evidence to a temporary analysis selection without changing the underlying ledger.
- Analyze and export. Optional AI receives only the selected snapshot and its coverage limitations. Reports retain the cited evidence and can be saved as a local PDF or portable dossier.
What works without AI
AI is not required to fetch or browse public records. The following remain deterministic and inspectable:
What AI changes—and what it does not
AI can summarize, compare, and propose evidence-linked interpretations. It does not fetch a hidden version of the address, silently expand the selected evidence, or turn a public label into verified identity.
The analyzed snapshot is frozen when the request starts. Records loaded later do not silently enter an in-flight report. Every completed or failed report preserves the context used at that moment.
Users choose the provider and model. Hosted-provider credentials and prompts travel directly from the browser to that provider, not through a Memoir AI proxy.
How to read the result
- Evidence is a returned onchain or public-index observation.
- Signal is a named source's statement or a reproducible metric over loaded evidence.
- Interpretation is an AI explanation that must cite supplied evidence or signal IDs.
- Coverage records what responded, what was partial, and what remains unknown.
Read the investigation methodology for the full distinction and coverage for current product limits.